Published October 25, 2026 | Version 2.0.0

A rapid and scalable observable for battery capacity loss detection enabled by symmetric bipolar pulse perturbations

  • 1. EDMO icon Chalmers University of Technology
  • 2. ROR icon Tsinghua University
  • 3. Xiamen Lijing New Energy Technology Co., Ltd.

Description

This dataset accompanies the manuscript “A rapid and scalable observable for battery capacity loss detection enabled by symmetric bipolar pulse perturbations.” It provides retired lithium-ion battery test data, processed features, and a reference model-output suite for studying electrochemical reversibility and battery capacity loss across cell chemistries.

Dataset contents and manuscript cohort

Version 2 covers 709 candidate cells across nine cell groups. Raw_Data.zip contains the raw workbooks for 708 cells. The raw workbook for 68 Ah LFP cell No. 6 (LFP_C_68_B_6_SOC_5-80_Part_1-1_ID_02HCB68111211Y88F0000101.xlsx) is supplied separately and completes the 709-cell raw-data inventory. After extracting Raw_Data.zip, place this workbook in Raw_Data/68Ah LFP/.

Data_Process_Output.zip contains the intermediate and final outputs of the four-step data-processing workflow for the 709 candidate cells, including Step4_SelectedFeatures_with_R0_Reff. The Step 4 workbooks include R0_avg and Reff_avg, calculated as the arithmetic means of the positive- and negative-pulse estimates at all five C-rates.

The cohort summarized in manuscript Table 1 comprises 673 cells after applying the documented SOH-based eligibility criteria. Individual analyses may contain fewer cells depending on the available test conditions. To reproduce the manuscript analyses, apply the cohort-selection procedure described in the code repository to the released processed data.

Model_Output.zip provides a reference modeling example for 20 Ah LFP cells at 40% SOC, a 5,000 ms pulse width, and 0.5C. It covers six feature configurations (fai_irrev, R0_avg, Reff_avg, raw_U, fai_irrev_R0_avg, and fai_irrev_Reff_avg), each evaluated with 100 random seeds and ten regression models.

Code availability

The corresponding data-processing and modeling code, with documentation and usage instructions, is publicly available at:

https://github.com/terencetaothucb/symmetric-bipolar-perturbations

Version history

Version 1: Initial public release of the raw data, processed outputs, and preliminary model examples.

Version 2: Added the separately supplied 68 Ah LFP cell No. 6 workbook, reissued the processed outputs with the unified Step 4 resistance features, and replaced the preliminary model examples with the six-configuration reference suite. Raw_Data.zip is unchanged from Version 1; the separately supplied workbook completes the raw-data record for a cell already represented in the processed outputs.

Files

Data_Process_Output.zip

Files (27.2 GB)

Name Size
md5:7f147bd2fa08648a1b9bea05ef7206a4
465.0 MB Preview Download
md5:9d057743b06d28f111e8728206c68dc9
77.0 MB Download
md5:5fbc84a244f5533230fc287345815d0e
52.6 MB Preview Download
md5:452e9ac71ec1173b32a53ab6b4997ff1
26.6 GB Preview Download

Additional details

Funding

Swedish Research Council
2023-04314
European Commission
Marie Skłodowska-Curie Actions 101283078
European Commission
Marie Skłodowska-Curie Actions 101131278
Tsinghua Shenzhen International Graduate School
Interdisciplinary Innovative Fund JC2021006